Re: [PATCH v3] mm/min_free_kbytes: modify min_free_kbytes calculation rules

From: Huang, Ying
Date: Sat Jun 24 2023 - 21:43:41 EST


Hi, Kefeng,

Kefeng Wang <wangkefeng.wang@xxxxxxxxxx> writes:

> Hi Liuq,
>
> On 2023/6/21 17:20, liuq wrote:
>> The current calculation of min_free_kbytes only uses ZONE_DMA and
>> ZONE_NORMAL pages,but the ZONE_MOVABLE zone->_watermark[WMARK_MIN]
>> will also divide part of min_free_kbytes.This will cause the min
>> watermark of ZONE_NORMAL to be too small in the presence of ZONE_MOVEABLE.
>> __GFP_HIGH and PF_MEMALLOC allocations usually don't need movable
>> zone pages, so just like ZONE_HIGHMEM, cap pages_min to a small
>> value in __setup_per_zone_wmarks.
>> On my testing machine with 16GB of memory (transparent hugepage is
>> turned off by default, and movablecore=12G is configured)
>> The following is a comparative test data of watermark_min
>> no patch add patch
>> ZONE_DMA 1 8
>> ZONE_DMA32 151 709
>> ZONE_NORMAL 233 1113
>> ZONE_MOVABLE 1434 128
>> min_free_kbytes 7288 7326
>>
>
> We see this issue and do the same change[1], and we add a per zone
> watermark configuration too, but both of them is not accepted,
> let's add Mel and wupeng to see more comments.
>
> [1]https://lore.kernel.org/linux-mm/20220905032858.1462927-1-mawupeng1@xxxxxxxxxx/

Thanks for your information! That's very useful.

As Mel said, ZONE_MOVABLE or ZONE_HIGHMEM pages allocation will need low
memory to be allocated too. But I don't understand why higher min
watermark of ZONE_MOVABLE will help this. It's too small compared with
the size of ZONE_MOVABLE in most cases. For example, on the system of
the patch description, the original "large" min watermark of
ZONE_MOVABLE is only (1434 * 4) = 5736 KB, which is too small compared
with 12 GB total ZONE_MOVABLE size. That is, before reaching min or low
watermark, we will allocate almost same number of ZONE_MOVABLE pages:
~12GB vs. ~(12GB - 5MB).

Best Regards,
Huang, Ying

>> Signed-off-by: liuq <liuq131@xxxxxxxxxxxxxxx>
>> ---
>> mm/page_alloc.c | 12 ++++++------
>> 1 file changed, 6 insertions(+), 6 deletions(-)
>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>> index 47421bedc12b..590ed8725e09 100644
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -6362,9 +6362,9 @@ static void __setup_per_zone_wmarks(void)
>> struct zone *zone;
>> unsigned long flags;
>> - /* Calculate total number of !ZONE_HIGHMEM pages */
>> + /* Calculate total number of !ZONE_HIGHMEM and !ZONE_MOVABLE pages */
>> for_each_zone(zone) {
>> - if (!is_highmem(zone))
>> + if (!is_highmem(zone) && zone_idx(zone) != ZONE_MOVABLE)
>> lowmem_pages += zone_managed_pages(zone);
>> }
>> @@ -6374,15 +6374,15 @@ static void __setup_per_zone_wmarks(void)
>> spin_lock_irqsave(&zone->lock, flags);
>> tmp = (u64)pages_min * zone_managed_pages(zone);
>> do_div(tmp, lowmem_pages);
>> - if (is_highmem(zone)) {
>> + if (is_highmem(zone) || zone_idx(zone) == ZONE_MOVABLE) {
>> /*
>> * __GFP_HIGH and PF_MEMALLOC allocations usually don't
>> - * need highmem pages, so cap pages_min to a small
>> - * value here.
>> + * need highmem and movable zones pages, so cap pages_min
>> + * to a small value here.
>> *
>> * The WMARK_HIGH-WMARK_LOW and (WMARK_LOW-WMARK_MIN)
>> * deltas control async page reclaim, and so should
>> - * not be capped for highmem.
>> + * not be capped for highmem and movable zones.
>> */
>> unsigned long min_pages;
>>